Data processing method and device, electronic device, and storage medium

Abstract

The disclosed embodiments are about a data processing method and device, an electronic device, and a storage medium, and relate to the field of artificial intelligence technology. The data processing method includes: dividing the data to be processed into labeled data and unlabeled data, labeling the unlabeled data according to the current model trained with the labeled data, and determining the selection strategy corresponding to the autonomous learning; selecting a target number of sample data from the unlabeled data using the selection strategy corresponding to the autonomous learning, and labeling the sample data to obtain labeled sample data to obtain updated labeled data; labeling the unlabeled data except the sample data according to the next model trained with the updated labeled data, until the termination condition is met to obtain the target model; predicting and labeling the remaining unlabeled data according to the target model to obtain the labeling result. The technical solution disclosed in the present invention can improve the accuracy and efficiency of data processing.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, and in particular to a data processing method, a data processing device, an electronic device, and a computer-readable storage medium. Background Art

[0002] In the current field of autonomous learning, the basic approach is to adopt a fixed selection strategy throughout the entire autonomous learning iteration process.

[0003] In the related art, selection strategies based on information volume and selection strategies based on diversity are generally adopted. Among them, the selection strategy based on information volume may extract similar samples, and there is a certain sampling bias, which makes the accuracy of the trained model poor. The calculation method of the selection strategy based on diversity has a large labor cost and wastes a lot of computing resources.

[0004] It should be noted that the information disclosed in the above background technology section is only used to enhance the understanding of the background of the present disclosure, and therefore may include information that does not constitute the prior art known to ordinary technicians in the field. Summary of the invention

[0005] The purpose of the present disclosure is to provide a data processing method and device, an electronic device, and a storage medium, thereby overcoming the problem of low accuracy caused by the limitations and defects of related technologies at least to a certain extent.

[0006] Other features and advantages of the present disclosure will become apparent from the following detailed description, or may be learned in part by the practice of the present disclosure.

[0007] According to one aspect of the present disclosure, a data processing method is provided, including: dividing the data to be processed into labeled data and unlabeled data, labeling the unlabeled data according to a current model trained using the labeled data, and determining a selection strategy corresponding to autonomous learning; the labeled data and the unlabeled data are used for autonomous learning; a target number of sample data are selected from the unlabeled data using the selection strategy corresponding to autonomous learning, and the sample data are labeled to obtain labeled sample data to obtain updated labeled data; according to a next model trained using the updated labeled data, the unlabeled data other than the sample data is labeled until a termination condition is met to obtain a target model; and the remaining unlabeled data are predictively labeled according to the target model to obtain a labeling result.

[0008] According to one aspect of the present disclosure, there is provided a data processing device, comprising: a selection strategy determination module, used to divide the data to be processed into labeled data and unlabeled data, label the unlabeled data according to a current model trained using the labeled data, and determine a selection strategy corresponding to autonomous learning; the labeled data and unlabeled data are used for autonomous learning; a sample selection module, used to select a target number of sample data from the unlabeled data using a selection strategy corresponding to autonomous learning, and label the sample data to obtain labeled sample data, so as to obtain updated labeled data; an iteration module, used to label the unlabeled data except the sample data according to a next model trained using the updated labeled data, until a termination condition is met to obtain a target model; a prediction and labeling module, used to predict and label the remaining unlabeled data according to the target model, and obtain a labeling result.

[0009] According to one aspect of the present disclosure, an electronic device is provided, comprising: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform any one of the above-mentioned data processing methods by executing the executable instructions.

[0010] According to one aspect of the present disclosure, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the data processing method described in any one of the above is implemented.

[0011] In the data processing method, data processing device, electronic device and computer-readable storage medium provided in the embodiments of the present disclosure, by determining the selection strategy corresponding to autonomous learning, and using the selection strategy corresponding to autonomous learning to select the target number of sample data from the unlabeled data pool, the labeled data is updated according to the labeled sample data, and the next model is continuously trained according to the updated labeled data, and the autonomous learning process is iterated in sequence until the target model is obtained, and the remaining unlabeled data is predicted and labeled by the target model. On the one hand, due to the different selection strategies corresponding to autonomous learning, the limitation of only one fixed selection strategy in the related art is avoided, and the selection strategy corresponding to each round of autonomous learning can be accurately determined, so that diverse sample data can be selected according to multiple selection strategies, and the accuracy of model training is improved. On the other hand, the appropriate selection strategy is automatically selected for each round of autonomous learning, and the user does not need to pay attention to the changes in each stage in a manual way to select the strategy, which improves the iterative process of autonomous learning, improves the labeling quality of data, and avoids the problem of occupying a large amount of computing resources when manually labeling data, reduces the computing resources and storage space required in the processing process, reduces the manual workload, simplifies the labeling steps, and thus improves the efficiency of data labeling and reduces the cost of data labeling.

[0012] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] The accompanying drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the specification are used to explain the principles of the present disclosure. Obviously, the accompanying drawings described below are only some embodiments of the present disclosure, and for ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without creative work.

[0014] Figure 1 A schematic diagram showing a system architecture to which the data processing method or data processing device according to the embodiment of the present disclosure can be applied.

[0015] Figure 2 A schematic diagram of the structure of an electronic device suitable for implementing the embodiments of the present disclosure is shown.

[0016] Figure 3 A schematic diagram schematically illustrates a data processing method in an embodiment of the present disclosure.

[0017] Figure 4 A schematic diagram for determining a selection strategy in an embodiment of the present disclosure is schematically shown.

[0018] Figure 5 The flowchart of the switching selection strategy in the embodiment of the present disclosure is schematically shown.

[0019] Figure 6 The overall flow chart of determining the selection strategy during the iterative process of autonomous learning in the embodiment of the present disclosure is schematically shown.

[0020] Figure 7 A flowchart schematically illustrates the first selection strategy sample data in an embodiment of the present disclosure.

[0021] Figure 8 A flowchart schematically illustrates sample data for the second selection strategy in an embodiment of the present disclosure.

[0022] Fig. 9 The overall flowchart of autonomous iteration in the embodiment of the present disclosure is schematically shown.

[0023] Fig.10 A schematic diagram schematically illustrates data in autonomous iteration in an embodiment of the present disclosure.

[0024] Fig.11 A schematic diagram schematically illustrates an application interface in an embodiment of the present disclosure.

[0025] Fig.12A block diagram of a data processing device in an embodiment of the present disclosure is schematically shown. DETAILED DESCRIPTION

[0026] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in a variety of forms and should not be construed as being limited to the examples set forth herein; on the contrary, these embodiments are provided so that the present disclosure will be more comprehensive and complete, and the concepts of the example embodiments are fully conveyed to those skilled in the art. The described features, structures, or characteristics may be combined in one or more embodiments in any suitable manner. In the following description, many specific details are provided to provide a full understanding of the embodiments of the present disclosure. However, those skilled in the art will appreciate that the technical solutions of the present disclosure may be practiced while omitting one or more of the specific details, or other methods, components, devices, steps, etc. may be adopted. In other cases, known technical solutions are not shown or described in detail to avoid obscuring various aspects of the present disclosure.

[0027] In addition, the accompanying drawings are only schematic illustrations of the present disclosure and are not necessarily drawn to scale. The same reference numerals in the figures represent the same or similar parts, and their repeated description will be omitted. Some of the block diagrams shown in the accompanying drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities can be implemented in software form, or implemented in one or more hardware modules or integrated circuits, or implemented in different networks and / or processor devices and / or microcontroller devices.

[0028] The present disclosure provides a data processing method. Figure 1 A schematic diagram showing a system architecture to which the data processing method or data processing device according to the embodiment of the present disclosure can be applied.

[0029] like Figure 1As shown, the system architecture 100 may include a first end 101, a network 102, and a second end 103. Among them, the first end may be a client, for example, a terminal such as a smart phone, a computer, a tablet computer, a smart speaker, etc. The network 102 is used to provide a medium for a communication link between the first end 101 and the second end 103. The network 102 may include various connection types, such as a wired communication link, a wireless communication link, etc. In the embodiment of the present disclosure, the network 102 between the first end 101 and the second end 103 may be a wired communication link, for example, a communication link may be provided through a serial port connection line, or a wireless communication link, and a communication link may be provided through a wireless network. The second end 103 may be a server or a client with a computing function, such as a portable computer, a desktop computer, a smart phone, etc., which is used to process the data sent by the first end. It should be noted that the first end may be the same as the second end, for example, it may be the same mobile terminal; the second end may also be different from the first end, for example, the first end is a mobile terminal and the second end is a server.

[0030] In an embodiment of the present disclosure, the first end can send the data to be processed to the second end, the second end divides the data to be processed, and uses the current model to perform and determine the selection strategy of autonomous learning on the unlabeled data; according to the selection strategy of autonomous learning, a target number of sample data are selected and labeled, and then the updated labeled data obtained by the labeled sample data is used to train the next model to label the unlabeled data other than the sample data until the target model is obtained, thereby predicting and labeling the remaining unlabeled data in the unlabeled data pool through the target model to obtain the labeling result.

[0031] It should be noted that the data processing method provided in the embodiment of the present disclosure can be completely executed by the client. Accordingly, the data processing device can be set in the client. It can also be executed by the server, which is not limited here.

[0032] Figure 2 Schematic diagram of an electronic device suitable for implementing an exemplary embodiment of the present disclosure is shown. The terminal of the present disclosure may be configured as follows Figure 2 The form of the electronic device shown, however, needs to be explained. Figure 2 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0033] The electronic device of the present disclosure includes at least a processor and a memory, wherein the memory is used to store one or more programs. When the one or more programs are executed by the processor, the processor can implement the method of the exemplary embodiment of the present disclosure.

[0034] Specifically, Figure 2As shown, the electronic device 200 may include: a processor 210, an internal memory 221, an external memory interface 222, a Universal Serial Bus (USB) interface 230, a charging management module 240, a power management module 241, a battery 242, an antenna 1, an antenna 2, a mobile communication module 250, a wireless communication module 260, an audio module 270, a speaker 271, a receiver 272, a microphone 273, an earphone interface 274, a sensor module 280, a display screen 290, a camera module 291, an indicator 292, a motor 293, a button 294, and a Subscriber Identification Module (SIM) card interface 295, etc. The sensor module 280 may include a depth sensor, a pressure sensor, a gyroscope sensor, an air pressure sensor, a magnetic sensor, an acceleration sensor, a distance sensor, a proximity light sensor, a fingerprint sensor, a temperature sensor, a touch sensor, an ambient light sensor, and a bone conduction sensor, etc.

[0035] It is to be understood that the structure illustrated in the embodiment of the present application does not constitute a specific limitation on the electronic device 200. In other embodiments of the present application, the electronic device 200 may include more or fewer components than shown in the figure, or combine some components, or split some components, or arrange the components differently. The components shown in the figure may be implemented in hardware, software, or a combination of software and hardware.

[0036] The processor 210 may include one or more processing units, for example: the processor 210 may include an application processor, a modem processor, a graphics processor, an image signal processor, a controller, a video codec, a digital signal processor, a baseband processor and / or a neural network processor (Neural-etwork Processing Unit, NPU), etc. Among them, different processing units can be independent devices or integrated in one or more processors. In addition, a memory can also be set in the processor 210 for storing instructions and data. The model training method in this exemplary embodiment can be executed by an application processor, a graphics processor or an image signal processor, and when the method involves neural network-related processing, it can be executed by an NPU.

[0037] The internal memory 221 can be used to store computer executable program codes, which include instructions. The internal memory 221 can include a program storage area and a data storage area. The external memory interface 222 can be used to connect an external memory card, such as a Micro SD card, to expand the storage capacity of the electronic device 200.

[0038] The communication function of the mobile terminal 200 can be implemented by a mobile communication module, an antenna 1, a wireless communication module, an antenna 2, a modulation and demodulation processor, and a baseband processor. Antenna 1 and antenna 2 are used to transmit and receive electromagnetic wave signals. The mobile communication module can provide 2G, 3G, 4G, 5G and other mobile communication solutions applied to the mobile terminal 200. The wireless communication module can provide wireless communication solutions such as wireless LAN, Bluetooth, and near field communication applied to the mobile terminal 200.

[0039] The display screen is used to implement display functions, such as displaying user interfaces, images, and videos. The camera module is used to implement shooting functions, such as shooting images and videos. The audio module is used to implement audio functions, such as playing audio and collecting voice. The power module is used to implement power management functions, such as charging the battery, powering the device, and monitoring the battery status.

[0040] The present application also provides a computer-readable storage medium, which may be included in the electronic device described in the above embodiment; or may exist independently without being assembled into the electronic device.

[0041] The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in conjunction with an instruction execution system, apparatus, or device.

[0042] Computer-readable storage media can send, propagate or transmit programs for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer-readable storage medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0043] The computer-readable storage medium carries one or more programs. When the one or more programs are executed by an electronic device, the electronic device implements the method described in the following embodiments.

[0044] The flow charts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flow chart or block diagram can represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flow chart, and the combination of the boxes in the block diagram or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0045] The units involved in the embodiments described in the present disclosure may be implemented by software or hardware, and the units described may also be arranged in a processor. The names of these units do not, in some cases, constitute limitations on the units themselves.

[0046] In the current AI scenarios, most of them are pure manual labeling modes, and some "algorithm pre-labeling" modes are adopted to improve efficiency (i.e., labeling with AI models first, and then manual verification based on model labeling). Both methods require manual labeling of all sample data. In order to save labeling costs, semi-supervised learning and autonomous learning methods have also emerged. Semi-supervised learning and autonomous learning methods are algorithms that are performed on a training data set that is a mixture of a small part of the labeled data set + a large part of the unlabeled data set. Semi-supervised learning methods mainly use clustering, similarity algorithms and other algorithms to attach pseudo-label information to unlabeled data to expand the labeled data set, but this pseudo-label information is directly obtained by the machine learning algorithm without manual calibration, which will introduce noise to the training of the AI ​​model and hinder the convergence of model training. The autonomous learning method introduces manual labeling experience into the training of the AI ​​model through each round of "model training-selection of unlabeled data-manual labeling" iteration until the stopping condition is met, which can steadily improve the model training effect. At the same time, when the iteration stops, a large amount of remaining unlabeled data is labeled with the model, which can save a lot of manual labeling costs. The core of autonomous learning is the selection strategy in each round of iteration. The purpose of the selection strategy is to select a batch of samples that are most conducive to improving the model training effect from the unlabeled data. There are already many selection strategies in the relevant technology, mainly information-based selection strategy and diversity-based selection strategy. The information-based selection strategy is also called uncertainty sampling, that is, selecting the samples that the current model considers to be the most uncertain. These samples contain rich information. Labeling such samples is most helpful to improve the current model. The information-based selection strategy is to find the samples with the highest information by all means. The diversity-based selection strategy selects samples from the information structure of unlabeled data, aiming to maximize the differences between the selected samples, avoid selecting similar samples, and make the selected samples represent the potential distribution of the entire data set as much as possible. The information-based selection strategy is the most commonly used autonomous learning selection strategy, but it has sampling bias and is likely to select a batch of very similar samples. Too many similar samples often cannot improve the model training effect enough in the early stage of autonomous learning. The diversity-based selection strategy can provide more data distribution information for model training in each round, improve the shortcomings of the information quantity-based selection strategy, and is more conducive to model training. It is especially suitable for the situation where there is little labeled training data in the early stage of autonomous learning. However, the diversity-based selection strategy often has a huge computational cost.

[0047] In the embodiment of the present disclosure, a data processing method is provided, which can be used in the data labeling process based on autonomous learning. Figure 3 The data processing method in the embodiment of the present disclosure is described in detail.

[0048] In step S310, the data to be processed is divided into labeled data and unlabeled data, the unlabeled data is labeled according to the current model trained using the labeled data, and a selection strategy corresponding to autonomous learning is determined; the labeled data and unlabeled data are used for autonomous learning.

[0049] In the embodiments of the present disclosure, it can be specifically applied to application scenarios of data labeling based on autonomous learning. The data to be processed refers to the data corresponding to the object to be processed. The object to be processed may be an image to be processed or an object of another type, etc. The labeled data refers to the data with labels obtained from all the data corresponding to the object to be processed, that is, the data of the labeled image. The labeled data may be data obtained by manually labeling part of the data to be processed, and the labeled data is used to realize autonomous learning of the model. The unlabeled data refers to the data that has not been labeled from all the data corresponding to the object to be processed, that is, the data of the unlabeled image. The unlabeled data in the unlabeled data pool is used for intelligent labeling through the trained model, and all the unlabeled data can form an unlabeled data pool for representing a collection of data that has not been manually labeled. Moreover, the number of labeled data is less than the number of unlabeled data contained in the unlabeled data pool.

[0050] Autonomous learning is used to obtain labeled data. The specific process may include: first, training the AI ​​model based on a small portion of labeled data, then obtaining unlabeled samples that contribute more to the target task based on the selection strategy, adding the labeled data after manual review and annotation to the next round of "model training-data selection-manual labeling" iteration until the pre-set stop condition is reached (such as the accuracy of each round of models on a fixed validation set, the proportion of manual labeling, etc.), and then using this round of models to label the remaining unlabeled data, in order to achieve the purpose of improving the model training effect and saving manual labeling costs.

[0051] In the disclosed embodiments, the current model refers to the current AI model, which is used to perform AI data annotation to determine the label of the data. AI data annotation is the process of annotating the corresponding data, such as image framing, image segmentation, text classification and slot extraction, etc.

[0052] After obtaining the labeled data, the model can be trained based on the labeled data to obtain the current model. For each round of autonomous learning, a current model is trained based on the labeled data of each round, and the parameters of the current model of each round can be different. Specifically, the labeled data and the corresponding true labels can be used as input to train the model, and the current model is trained by comparing the predicted labels with the true labels. The labeled data of each round can be updated in real time.

[0053] All unlabeled data in the unlabeled data pool can be labeled based on the current model to obtain the attribute information of each unlabeled data under the current model. The attribute information may include information volume, which includes but is not limited to methods based on classification confidence, margin, entropy, etc. Further, the selection strategy for multiple stages can be determined based on the information volume using an automatic staged selection strategy. The purpose of the selection strategy is to select a batch of samples from the unlabeled data that are most conducive to improving the model training effect.

[0054] Autonomous learning can be divided into multiple different stages, and each stage can include one or more rounds of autonomous learning processes. Based on this, the selection strategy corresponding to autonomous learning can be automatically determined by the automatic staged selection strategy to divide all rounds of autonomous learning processes into the first stage and the second stage. The first stage and the second stage are divided according to whether the staged conditions are met and the index parameter conditions are met. Meeting the staged conditions refers to the stage where the next selection strategy can be entered, and meeting the index parameter conditions refers to the index parameter being less than 10. For example, if the staged conditions are not met or the index parameter conditions are not met, it belongs to the first stage, and if the staged conditions and the index parameter conditions are met, it belongs to the second stage. The first stage can be the early stage, including all rounds of autonomous learning from the starting moment to the staged conditions and the index parameter conditions are met, and the second stage can be the later stage, including all rounds of autonomous learning after the first stage. The first stage and the second stage can represent multiple rounds of autonomous learning respectively, and the number of rounds of autonomous learning represented by the first stage and the second stage can be the same or different.

[0055] If the autonomous learning belongs to different stages, the corresponding selection strategies are different. For example, the first stage can be a mixed selection strategy, and the second stage can be a selection stage based on the amount of information. Based on this, in order to accurately extract samples, the selection strategy corresponding to each round of autonomous learning can be determined.

[0056] Figure 4 A flow chart for determining the selection strategy is schematically shown in Figure 4 As shown, it mainly includes the following steps:

[0057] In step S410, the first selection strategy is determined as the selection strategy corresponding to the current round of autonomous learning;

[0058] In step S420, if it is determined that the stage conditions are met, the selection strategy corresponding to the next round of autonomous learning is determined by combining the first selection strategy and the second selection strategy.

[0059] In the embodiment of the present disclosure, in the first stage, the first selection strategy can be directly used as the selection strategy for the current round of autonomous learning, and the current round of autonomous learning can be any round in the entire autonomous learning process. The first selection strategy can be a hybrid selection strategy. The hybrid selection strategy refers to a selection strategy based on the amount of information and a selection strategy based on diversity. N sample data, that is, the target number of sample data, can be selected from the unlabeled data pool according to the first selection strategy.

[0060] During the use of the first selection strategy, it can be determined whether the current round of autonomous learning meets the phased conditions. Phased conditions are used to describe whether the current round of autonomous learning can enter the second phase. Whether the current round of autonomous learning meets the phased conditions can be determined based on the validation set indicators and the attribute information of multiple sample data selected according to the first selection strategy. The data in the validation set has labels, and the validation set is used to monitor the training effect of each round of the model. The validation set indicator refers to the accuracy of the model on the validation set.

[0061] Among them, if the validation set indicators for multiple consecutive rounds have not increased, or the attribute information of the target number of sample data selected according to the first selection strategy has not increased relative to the attribute information of the previous round, it is determined that the stage-by-stage conditions are met. Specifically, based on the comparison results of the information volume of the multiple sample data selected by the first selection strategy in the current round with the multiple sample data corresponding to the previous round, or based on the judgment results of the validation set indicators of the current model for the validation set, it is determined whether the stage-by-stage conditions are met. If the information volume variance of the selected N sample data has not increased (not increased) compared with the N sample data in the previous round, or the accuracy of the current model on the validation set has not increased (not increased) for two consecutive rounds, it is determined that the stage-by-stage conditions are met. In order to better judge whether autonomous learning can enter the next selection strategy, that is, the second selection strategy, the information volume variance of the N sample data selected by the rth round of autonomous learning can be shown as formula (1):

[0062]

[0063] Among them, φ i is the information content of the i-th sample data,

[0064] If it is determined that the current round of autonomous learning does not meet the stage conditions, the first selection strategy will continue to be executed to perform the next round of autonomous learning iteration. If the stage conditions are met, the control will enter the second stage of autonomous learning from the first stage of autonomous learning, and the first selection strategy will be switched to the second selection strategy to determine the selection strategy for the next round of autonomous learning in combination with the first selection strategy and the second selection strategy. The second selection strategy can be a selection strategy based on the amount of information. Starting from the next round, each subsequent round of autonomous learning can gradually transition from the first selection strategy to the second selection strategy, that is, the first selection strategy is first executed, and then the first selection strategy is adjusted to switch the first selection strategy to the second selection strategy.

[0065] When the first selection strategy is switched to the second selection strategy, the switch can be made according to the comparison result between the indicator parameter and the parameter threshold. The indicator parameter refers to the data set expansion index. The data set expansion index B refers to the parameter of the hybrid selection strategy adopted in the current round, which can be determined according to formula (2), where N is the pre-set amount of unlabeled data selected in each round of autonomous learning, that is, the target amount. The amount of unlabeled data in the current round / / N refers to the integer after the division operation.

[0066] B = min (the amount of unlabeled data in the current round / / N, 10) Formula (2)

[0067] Figure 5 A flow chart of the switching selection strategy is shown in Figure 5 As shown in , it mainly includes the following steps:

[0068] In step S510, it is determined whether the indicator parameter is less than the parameter threshold. If so, the process proceeds to step S520; if not, the process proceeds to step S530.

[0069] In this step, the parameter threshold can be determined in advance according to actual needs. The parameter threshold can be, for example, 10 or other values. Here, 10 is used as an example for illustration. The indicator parameter can be compared with the parameter threshold to obtain a comparison result that the indicator parameter is greater than the parameter threshold or the indicator parameter is less than the parameter threshold. Further, the first selection strategy can be switched to the second selection strategy according to the comparison result.

[0070] In step S520, if the indicator parameter in the first selection strategy is less than the parameter threshold, the first selection strategy is directly switched to the second selection strategy.

[0071] In this step, if the data set expansion index is less than 10, the first selection strategy can be directly converted to the second selection strategy, and each subsequent round of autonomous learning adopts a selection strategy based only on the amount of information.

[0072] In step S530, if the indicator parameter is not less than the parameter threshold, the indicator parameter of the first selection strategy is adjusted, and the selection strategy is converted from the first selection strategy to the second selection strategy.

[0073] In this step, if the data set expansion index is not less than 10, the indicator parameter can be decayed to gradually transition the first selection strategy to the second selection strategy through the reference selection strategy. The reference selection strategy refers to the first selection strategy with dynamic changes in the data set expansion index. When decaying the indicator parameter, the indicator parameter is first reduced, and then the reduced indicator parameter is decreased. When reducing, the indicator parameter is reduced according to the first fixed parameter, and the first fixed parameter can be 2. When decreasing, it can be decreased according to the second fixed parameter. The second fixed parameter can be 1, and the decay can be stopped when the indicator parameter decays to the reference indicator parameter. The reference indicator parameter can be 1. That is, the indicator parameter B is decayed by 1 / 2 to B / 2, and then the indicator parameter B is reduced by 1 in each round of autonomous learning of the mini-batch kmeans hybrid selection strategy until the indicator parameter B is equal to 1 and stops decaying. Mini-batch kmeans clustering, first the total data set is screened according to some conditions to obtain a smaller mini-batch data, and then the kmeans clustering operation is performed on this mini-batch. When the indicator parameter decays to 1, the first selection strategy is switched to the second selection strategy. Therefore, it can be understood that if the indicator parameter is not less than the parameter threshold, the first selection strategy is first adjusted to continue the next round of autonomous learning according to the mixed selection strategy, until the indicator parameter B is equal to 1, the decay stops, and each subsequent round of autonomous learning is performed according to the single selection strategy represented by the second selection strategy.

[0074] Figure 6 A complete autonomous learning iterative process to determine the specific flow chart of the selection strategy is shown schematically in Figure 6 As shown in , it mainly includes the following steps:

[0075] Step S610, determining the indicator parameter B of the hybrid selection strategy used in the current round. The hybrid selection strategy is the first selection strategy. Since the amount of unlabeled data in each round is different, the indicator parameters corresponding to the hybrid selection strategy in each round of autonomous learning are different.

[0076] Step S620: This round of autonomous learning adopts a mixed selection strategy with parameter B.

[0077] Step S630: select N selected samples from the unlabeled data pool. The N selected samples refer to N sample data.

[0078] Step S640, determine whether the current round of autonomous learning meets the stage conditions; if not, repeat the autonomous learning iterations from step S610 to step S630; if so, execute step S650.

[0079] Step S650, determine whether the indicator parameter B at this time is less than the parameter threshold; if so, execute step S670; if not, execute step S660.

[0080] Step S660, the indicator parameter B is attenuated by 1 / 2 to B / 2, and then B is reduced by 1 in each round of hybrid selection strategy of autonomous learning until the indicator parameter B is equal to 1. That is, the indicator parameter of the first selection strategy is attenuated in each round of autonomous learning starting from the next round of autonomous learning.

[0081] Step S670: Each round of autonomous learning thereafter adopts a selection strategy based only on the amount of information, that is, each round of autonomous learning starting from the next round of autonomous learning adopts the second selection strategy. Or each round of autonomous learning starting from the autonomous learning when B is equal to 1 adopts the second selection strategy.

[0082] In the disclosed embodiment, for each round of autonomous learning, the first selection strategy is first performed and it is automatically determined whether to switch to the second selection strategy. By determining the selection strategy corresponding to each stage of autonomous learning, rather than using a fixed selection strategy throughout the entire autonomous learning iteration process, the limitations and singleness of using only one selection strategy are avoided, and the accuracy and rationality are improved. Based on this, the accuracy of the joint validation set and the variance of the information volume of the selected data are monitored to automatically determine whether the autonomous learning should enter the next selection strategy stage, so that each round of autonomous learning adopts the most appropriate selection strategy to accelerate the autonomous learning iteration process, improve the model training effect and the final intelligent labeling quality. In addition, it can automatically identify whether the current round of autonomous learning belongs to the first stage or the second stage, and automatically determine whether it is the next selection strategy stage, and automatically determine whether to switch the selection strategy, avoiding possible misoperations caused by manual judgment, and can improve the operation efficiency and the accuracy of autonomous learning.

[0083] In step S320, a target number of sample data is selected from the unlabeled data using a selection strategy corresponding to autonomous learning, and the sample data is labeled to obtain labeled sample data, so as to obtain updated labeled data.

[0084] In the disclosed embodiment, a selection strategy corresponding to autonomous learning can be used to select a target number of sample data from an unlabeled data pool. That is, in the first stage, a target number of sample data is selected according to a first selection strategy, and in the second stage, a target number of sample data is selected according to a second selection strategy. After obtaining the target number of sample data, the sample data can be labeled to obtain labeled sample data. The labeling process here can be manual labeling. Furthermore, the labeled data can be updated based on the labeled sample data to obtain updated labeled data. Specifically, the labeled sample data can be added to the labeled data set to obtain updated labeled data. It should be noted that after each round of iteration is completed, the labeled data can be updated based on the sample data manually labeled in that round.

[0085] Among them, when selecting unlabeled data according to the first selection strategy, the specific steps include: predicting the unlabeled data in the unlabeled data pool through the current model to obtain a prediction score, and calculating the amount of information of the unlabeled data based on the prediction score. The prediction score can be a prediction confidence. The amount of information may include but is not limited to classification confidence, confidence margin, and confidence entropy. Further, a reference number of sample data is obtained according to the arrangement order of the amount of information; the reference number may be greater than the target number and is determined according to the indicator parameter and the target number. For example, it can be the product of the indicator parameter and the target number B*N, where B is the data set expansion index, and N is the pre-set unlabeled data selected in each round of autonomous learning, that is, the target number.

[0086] On this basis, the sample data of the reference quantity can be clustered to obtain the sample data of the target quantity. The sample data of the reference quantity can be clustered using a clustering algorithm to obtain the cluster centers of the target quantity to obtain the sample data of the target quantity. The clustering algorithm may include but is not limited to kmeans clustering, spectral clustering, and hierarchical clustering, etc., and kmeans clustering is used as an example for explanation. Specifically, among the sample data of the reference quantity, a sample data closest to the cluster center is selected to obtain the sample data of the target quantity.

[0087] Figure 7 The flowchart of selecting sample data by the first selection strategy is schematically shown in FIG. Figure 7 As shown in , it mainly includes the following steps:

[0088] Step S701, the current model obtained in the current round of training predicts unlabeled data.

[0089] Step S702: The current model predicts the unlabeled data to obtain a prediction score, and calculates the amount of information of the unlabeled data based on the prediction score.

[0090] Step S703, selecting the top B*N sample data in terms of information volume.

[0091] Step S704, select N samples that are closest to the corresponding cluster centers. Perform kmeans clustering on B*N sample data and divide them into N categories, and obtain N cluster centers.

[0092] Step S705, perform manual labeling and add the labeled data.

[0093] Among them, the clustering features can use the deep network coding of text, the pixel information matrix of the image, etc. The specific algorithm steps of kmeans are:

[0094] 1. Select the initialized N samples as the initial cluster centers {a1, a2, ..., a N};

[0095] 2. For each sample x in the data set i Calculate its distance to N cluster centers and divide it into the class corresponding to the cluster center with the smallest distance;

[0096] 3. Use {c1,c2,...,c N} represents the N categories divided into, and the corresponding cluster centers are {a1,a2,...,a N}, for each category c j , recalculate its cluster center according to formula (3), where |c j |Marking Category c j Amount of data contained in:

[0097]

[0098] 4. Repeat steps 2 and 3 above until the end condition is reached. The end condition can be determined based on the number of iterations being greater than the preset number or the minimum error change being less than the preset change value.

[0099] 5. Among the B*N sample data divided into N categories, select a sample data closest to the cluster center, and N categories can get N sample data;

[0100] 6. After manually annotating the N sample data, add them to the annotated data set and prepare for the next round of autonomous learning iteration.

[0101] For the selection strategy in the early stage of autonomous learning, in order to take advantage of the diversity selection strategy while overcoming its disadvantage of high computational cost, a mini-batch kmeans hybrid selection strategy is adopted. The hybrid strategy can take into account the advantages of both information volume and diversity selection strategy.

[0102] Specifically, when selecting unlabeled data according to the second selection strategy, the specific steps include: predicting the unlabeled data in the unlabeled data pool through the current model to obtain a prediction score. The prediction score can be a prediction confidence. Taking the classification task as an example, assuming that the total label set is M = {y1, y2, ..., y M}, the total number of labels is m, and the prediction confidence of a data x is usually a vector vector The dimension is the total number of categories m, The expression of is shown in formula (4), where p(y i |x),x=1,2,,...,m means sample x is predicted to be label category y i The classification confidence of:

[0103]

[0104] When intelligently labeling a model, the label category corresponding to the maximum confidence level is usually selected, as shown in formula (5):

[0105]

[0106] The information volume is further calculated based on the prediction score of the unlabeled data, and the target number of sample data is obtained from the unlabeled data pool according to the arrangement order of the information volume. There are many calculation indicators for information volume, usually based on classification confidence, margin, entropy and other methods. Assume that U represents the unlabeled data set and the total label set is M = {y1, y2, ..., y M}, the amount of information of each type can be calculated according to the corresponding method.

[0107] The amount of information calculated based on classification confidence is shown in formula (6):

[0108]

[0109] The margin-based selection strategy is shown in formula (7):

[0110]

[0111] in, and Respectively represent the two maximum classification confidences of the data x.

[0112] The entropy-based selection strategy is shown in formula (8):

[0113]

[0114] Next, we can select the sample data with the first N information amounts from the unlabeled data pool in descending order as the target number of sample data. N The sample data representing the target number selected can be selected by the formula (9):

[0115] X N =arg topN max x∈U φ x Formula (9)

[0116] where arg topN Indicates selecting the sample data whose information content meets the conditions and ranks in the top N.

[0117] Figure 8 A flow chart schematically shows the second selection strategy, see Figure 8 As shown in , it mainly includes the following steps:

[0118] In step S801, the current model obtained in the current round of training predicts the unlabeled data pool;

[0119] In step S802, the current model predicts the unlabeled data to obtain a prediction score; based on the prediction score of the unlabeled data, the amount of information is calculated;

[0120] In step S803, the top N sample data in terms of information volume are selected from the unlabeled data pool;

[0121] In step S804, the remaining samples continue to remain in the unlabeled data pool, and the selected samples are manually labeled and added to the labeled data set, preparing for the next round of autonomous learning iteration.

[0122] In the disclosed embodiment, the selection strategy is gradually transitioned to one based on the amount of information by changing the parameter B of the mini-batch kmeans in the later stage, which can also ensure that the model training effect in the middle and late stages of autonomous learning steadily increases.

[0123] In summary, the technical solution provided in the embodiment of the present disclosure, through the staged change selection strategy, uses the mini-batch kmeans mixed selection strategy of diversity and information volume in the early stage of autonomous learning iteration, which can select richer and more diverse samples, so that the model training effect is better and the accuracy of the verification set is improved faster. At the same time, it can reduce the computational overhead and improve the iteration efficiency. In the later stage, the selection strategy is gradually transitioned to the information volume-based one through the change of the parameter B of the mini-batch kmeans, which can ensure that the model training effect in the middle and late stages of autonomous learning steadily increases. The staged change selection strategy of the method does not require manual operation, but is combined and automatically determined by the change of the information volume variance of each round of sample data and the accuracy of the verification set to determine whether autonomous learning should enter the next selection strategy stage, so as to avoid manual misjudgment and operation affecting the effect of autonomous learning. Based on this, determining the selection strategy for each stage can accelerate the autonomous learning iteration process, improve the accuracy of the labeled data and the accuracy of the model, and the whole process is intelligent and efficient, which can save a considerable amount of labor costs for the current AI data labeling and provide the necessary basic conditions for the rapid development of artificial intelligence.

[0124] After determining the selection strategy for autonomous learning through an automatic phased selection strategy, a target number of sample data can be obtained from the unlabeled data according to the selection strategy corresponding to each round of autonomous learning, and then these sample data can be manually labeled to obtain labeled sample data. This part of the labeled sample data is removed from the unlabeled data pool and added to the labeled data for the next round of autonomous learning.

[0125] In step S330, the unlabeled data other than the sample data is labeled according to the next model trained using the updated labeled data, until the termination condition is met to obtain the target model.

[0126] In the disclosed embodiment, the labeled data can be updated based on the labeled sample data to obtain updated labeled data. Then, the model is trained based on the updated labeled data to obtain the next model. Then, the data in the unlabeled data pool other than the labeled sample data is labeled based on the next model to obtain the amount of information of each data under the next model. According to the amount of information, an automatic phased selection strategy is used to select a target number of sample data from the data, and the target number of sample data is further manually labeled, and updated labeled data is obtained again, so that the target number of sample data can maximize the effect of the next round of model training after obtaining the manually labeled labels.

[0127] Each autonomous learning process includes training model-data selection-manual labeling. Based on this, the above-mentioned training model, labeling the remaining unlabeled data through the model to calculate the amount of information, and extracting the target number of sample data and labeling by staged selection strategies can be repeated until the termination condition is met to complete the entire autonomous learning iteration process. Among them, the labeling parameters of each model can be monitored according to the verification set, and the termination condition is determined to be met when the labeling parameters meet the labeling condition. Each model refers to the model trained in each round of autonomous iteration. The labeling parameters are the accuracy of the verification set and the proportion of manually labeled data. Based on this, the termination condition can be considered to be met when the accuracy of the verification set is greater than the preset value and / or the proportion of manually labeled data reaches the preset proportion. When autonomous learning reaches the termination condition, the iteration is stopped and the target model is obtained. The target model refers to the latest model obtained after multiple rounds of autonomous learning.

[0128] In step S340, the remaining unlabeled data is predicted and labeled according to the target model to obtain a labeling result.

[0129] In this step, after obtaining the target model, the remaining unlabeled data in the unlabeled data pool can be labeled according to the target model to obtain the corresponding labeling results, so as to obtain the largest scale of labeled data and the optimal target model. Further, the target operation can be performed according to the labeling results. The target operation can be determined according to the actual application scenario, for example, it can be an identification operation or a classification operation. Here, the target operation is described as an identification operation.

[0130] Fig. 9 The overall process flow diagram is schematically shown. Fig. 9 As shown in , it mainly includes the following steps:

[0131] Step S901, dividing the data to be processed into a small amount of labeled data and an unlabeled data pool.

[0132] Step S902: Perform AI model training based on the labeled data.

[0133] Step S903: Use the trained model to predict the unlabeled data to obtain the information content of each unlabeled data under the current model prediction.

[0134] Step S904: Use an automatic phased selection strategy to select a target number of sample data from the unlabeled data pool.

[0135] Step S905 , manually label the target number of sample data, remove them from the unlabeled data pool and add them to the labeled data.

[0136] Step S906, repeat steps S902 to S905, each repetition is an autonomous learning iteration, and there will be an independent labeled validation set to monitor the effect of each round of model training. The stopping condition can be pre-set to control the iteration process, and the stopping condition can be the accuracy of the model on the validation set, the proportion of manually labeled data, etc.

[0137] Step S907, when the autonomous learning reaches the termination condition, the iteration is stopped, and the target model is used to predict and label the remaining unlabeled data.

[0138] On this basis, the data to be processed corresponding to the object to be processed can be divided into labeled data and unlabeled data, and the unlabeled data can be labeled according to the current model trained using the labeled data, and the selection strategy corresponding to the autonomous learning can be determined; the target number of sample data can be selected from the unlabeled data using the selection strategy corresponding to the autonomous learning, and the sample data can be labeled to obtain labeled sample data to obtain updated labeled data; according to the next model trained using the updated labeled data, the unlabeled data other than the sample data can be labeled until the termination condition is met to obtain the target model; the remaining unlabeled data can be predicted and labeled according to the target model to obtain the labeling result. Further, the target operation can be performed on the object to be processed according to the labeling result to obtain the processing result corresponding to the target operation. For example, the type of the image to be processed can be identified.

[0139] Fig.10 Schematically shows the data processing process in the autonomous iteration process, dividing the data to be processed 1000 into labeled data 1001 and unlabeled data 1002. According to the selection strategy corresponding to the autonomous learning, a target number of sample data 1003 is selected from the unlabeled data, and the target number of sample data is added to the labeled data. This cycle is repeated until the termination condition is met to obtain the remaining unlabeled data 1004.

[0140] The technical solution of the disclosed embodiment provides an efficient and intelligent autonomous learning system for data labeling. The user only needs to provide a small amount of manually labeled data in the initial stage for autonomous model learning. After the model is trained, the unlabeled data is intelligently labeled, and an automatic phased selection strategy is used to screen out some samples for manual verification and labeling. The verified data is used for the next round of autonomous learning of the model. After the model is trained, the next round of intelligent labeling is started until the preset termination condition is met. This solution enables the user to ultimately obtain large-scale labeled data and the optimal model.

[0141] When using the above method for data annotation, you can set the specific information volume indicators used in the phased selection strategy, the amount of data selected for manual verification in each round, and the iterative termination conditions for autonomous learning. Fig.11As shown in the initial configuration interface, the selection strategy is based on the amount of information, the amount of data selected in each round is 50, and the autonomous learning iteration termination condition is the model accuracy. When the autonomous learning stops iterating, the target model intelligently labels the remaining unlabeled data to obtain high-quality large-scale labeled data and the optimal target model.

[0142] The technical solution proposed in the embodiment of the present disclosure adopts a mini-batch kmeans hybrid selection strategy that takes into account diversity and information volume in the initial stage of autonomous learning iteration, i.e., the first stage, so that the model training effect is rapidly improved. And through the information volume variance of the sample data, the information volume variance of the sample data and the accuracy of the verification set are monitored in each round of autonomous learning, and the parameters in the mini-batch kmeans hybrid selection strategy are automatically changed, so that it gradually transitions to a selection strategy based on information volume, so that the model performance effect is steadily improved. The technical solution provided by the present disclosure does not require the user to pay attention to the changes in each stage, and automatically implements the staged change selection strategy, reducing the cost of manual annotation during the autonomous learning process.

[0143] The present disclosure provides a data processing device, referring to Fig.12 As shown in , the data processing device 1200 may include:

[0144] A selection strategy determination module 1201 is used to divide the data to be processed into labeled data and unlabeled data, label the unlabeled data according to the current model trained using the labeled data, and determine the selection strategy corresponding to the autonomous learning; the labeled data and the unlabeled data are used for autonomous learning;

[0145] The sample selection module 1202 is used to select a target number of sample data from the unlabeled data using a selection strategy corresponding to the autonomous learning, and label the sample data to obtain labeled sample data, so as to obtain updated labeled data;

[0146] Iteration module 1203, used for labeling the unlabeled data except the sample data according to the next model trained by using the updated labeled data, until the termination condition is met to obtain the target model;

[0147] The prediction and labeling module 1204 is used to predict and label the remaining unlabeled data according to the target model to obtain labeling results.

[0148] In an exemplary embodiment of the present disclosure, the selection strategy determination module includes: a current round determination module, used to determine the first selection strategy as the selection strategy corresponding to the current round of autonomous learning; a next round determination module, used to determine the selection strategy corresponding to the next round of autonomous learning based on the verification set indicators and the attribute information of each unlabeled data under the current model; the attribute information is obtained by labeling the unlabeled data according to the current model.

[0149] In an exemplary embodiment of the present disclosure, the next round determination module includes: a first determination module, which is used to determine the first selection strategy as the selection strategy corresponding to the next round of autonomous learning if it is determined that the current autonomous learning does not meet the stage conditions based on the verification set indicators and the attribute information; a second determination module, which is used to determine the selection strategy corresponding to the next round of autonomous learning by combining the first selection strategy and the second selection strategy if it is determined that the current autonomous learning meets the stage conditions.

[0150] In an exemplary embodiment of the present disclosure, the device also includes: a conditional judgment module, which is used to determine that the stage-by-stage conditions are met if the verification set indicators of multiple consecutive rounds have not increased, or the attribute information of the target number of sample data selected according to the first selection strategy has not increased compared with the attribute information of the previous round.

[0151] In an exemplary embodiment of the present disclosure, the second determination module includes: a switching module, which is used to directly switch the first selection strategy to the second selection strategy if the indicator parameter in the first selection strategy is less than the parameter threshold; and a transition module, which is used to adjust the indicator parameter of the first selection strategy if the indicator parameter is not less than the parameter threshold, and transition the selection strategy from the first selection strategy to the second selection strategy.

[0152] In an exemplary embodiment of the present disclosure, the transition module includes: a decay module, configured to decay the indicator parameter until the indicator parameter decays to a reference indicator parameter and then transition the first selection strategy to the second selection strategy.

[0153] In an exemplary embodiment of the present disclosure, the sample selection module includes: an information quantity calculation module, which is used to predict the unlabeled data in the unlabeled data pool through the current model to obtain a prediction score, and calculate the information quantity of the unlabeled data based on the prediction score; a sample acquisition module, which is used to obtain a reference number of sample data according to the arrangement order of the information quantity; and a sample screening module, which is used to cluster the reference number of sample data to obtain a target number of sample data.

[0154] In an exemplary embodiment of the present disclosure, the sample screening module includes: a clustering module, which is used to select a sample data closest to the cluster center from the reference number of samples to obtain the target number of sample data; the reference number of sample data corresponds to the category of the target number.

[0155] In an exemplary embodiment of the present disclosure, the sample selection module includes: a prediction score determination module, which is used to predict the unlabeled data through the current model to obtain the prediction score; a selection control module, which is used to calculate the amount of information based on the prediction score of the unlabeled data, and obtain the target number of sample data according to the arrangement order of the information amount.

[0156] It should be noted that the specific details of each module in the above-mentioned data processing device have been described in detail in the corresponding data processing method, so they will not be repeated here.

[0157] Through the description of the above implementation, it is easy for those skilled in the art to understand that the example implementation described here can be implemented by software, or by software combined with necessary hardware. Therefore, the technical solution according to the implementation of the present disclosure can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.) or on a network, including several instructions to enable a computing device (which can be a personal computer, a server, a terminal device, or a network device, etc.) to execute the method according to the implementation of the present disclosure.

[0158] In addition, the above-mentioned figures are only schematic illustrations of the processes included in the method according to the exemplary embodiments of the present disclosure, and are not intended to be limiting. It is easy to understand that the processes shown in the above-mentioned figures do not indicate or limit the time sequence of these processes. In addition, it is also easy to understand that these processes can be performed synchronously or asynchronously, for example, in multiple modules.

[0159] It should be noted that, although several modules or units of the device for action execution are mentioned in the above detailed description, this division is not mandatory. In fact, according to the embodiments of the present disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. On the contrary, the features and functions of one module or unit described above can be further divided into multiple modules or units to be embodied.

[0160] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing what is disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary technical means in the art that are not disclosed in the present disclosure. The specification and embodiments are to be considered merely as exemplary, and the true scope and spirit of the present disclosure are indicated by the claims.

[0161] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A data processing method, characterized in that: include: Divide the data to be processed into labeled data and unlabeled data, label the unlabeled data according to the current model trained using the labeled data, and determine the selection strategy corresponding to the autonomous learning; the labeled data and the unlabeled data are used for autonomous learning; wherein the data to be processed is image data; Selecting a target number of sample data from unlabeled data using a selection strategy corresponding to autonomous learning, and labeling the sample data to obtain labeled sample data, so as to obtain updated labeled data; According to the next model trained with the updated labeled data, unlabeled data other than the sample data is labeled until a termination condition is met to obtain a target model; Predictively label the remaining unlabeled data according to the target model to obtain labeling results; Wherein, the determination of the selection strategy corresponding to the autonomous learning includes: Determine the first selection strategy as the selection strategy corresponding to the current round of autonomous learning; According to the validation set indicators and the attribute information of each unlabeled data under the current model, the selection strategy corresponding to the next round of autonomous learning is determined; the attribute information is obtained by labeling the unlabeled data according to the current model; wherein, if it is determined according to the validation set indicators and the attribute information that the current autonomous learning does not meet the stage conditions, the first selection strategy is determined as the selection strategy corresponding to the next round of autonomous learning; if it is determined that the current autonomous learning meets the stage conditions, the selection strategy corresponding to the next round of autonomous learning is determined by combining the first selection strategy and the second selection strategy.

2. The data processing method according to claim 1, characterized in that: The method further comprises: If the validation set index does not increase for consecutive rounds, or the attribute information of the target number of sample data selected according to the first selection strategy does not increase relative to the attribute information of the previous round, it is determined that the stage-by-stage conditions are met.

3. The data processing method according to claim 1, characterized in that: The combining the first selection strategy and the second selection strategy to determine the selection strategy corresponding to the next round of autonomous learning includes: If the indicator parameter in the first selection strategy is less than the parameter threshold, directly switch the first selection strategy to the second selection strategy; If the indicator parameter is not less than the parameter threshold, the indicator parameter of the first selection strategy is adjusted to transition the selection strategy from the first selection strategy to the second selection strategy, including: attenuating the indicator parameter until the indicator parameter decays to a reference indicator parameter and then transitioning the first selection strategy to the second selection strategy.

4. The data processing method according to claim 1, characterized in that: The method of using the selection strategy corresponding to autonomous learning to select a target number of sample data from unlabeled data includes: Predicting the unlabeled data in the unlabeled data pool by the current model to obtain a prediction score, and calculating the amount of information of the unlabeled data based on the prediction score; Acquire a reference amount of sample data according to the arrangement order of the information amount; The reference amount of sample data is clustered to obtain a target amount of sample data.

5. The data processing method according to claim 4, characterized in that: The clustering of the reference amount of sample data to obtain the target amount of sample data includes: Among the sample data of the reference number, a sample data closest to the cluster center is selected respectively to obtain the sample data of the target number; the sample data of the reference number corresponds to the category of the target number.

6. The data processing method according to claim 1, characterized in that: The method of using the selection strategy corresponding to autonomous learning to select a target number of sample data from unlabeled data includes: Use the current model to predict unlabeled data and obtain prediction scores; The amount of information is calculated based on the prediction score of the unlabeled data, and a target number of sample data is obtained according to the arrangement order of the amount of information.

7. A data processing device, characterized in that: include: A selection strategy determination module is used to divide the data to be processed into labeled data and unlabeled data, label the unlabeled data according to the current model trained using the labeled data, and determine the selection strategy corresponding to the autonomous learning; the labeled data and the unlabeled data are used for autonomous learning; wherein the data to be processed is image data; A sample selection module is used to select a target number of sample data from unlabeled data using a selection strategy corresponding to autonomous learning, and to label the sample data to obtain labeled sample data, so as to obtain updated labeled data; An iteration module, used for labeling the unlabeled data other than the sample data according to the next model trained by using the updated labeled data, until a termination condition is met to obtain a target model; A prediction and labeling module is used to predict and label the remaining unlabeled data according to the target model to obtain labeling results; Wherein, the determination of the selection strategy corresponding to the autonomous learning includes: Determine the first selection strategy as the selection strategy corresponding to the current round of autonomous learning; According to the validation set indicators and the attribute information of each unlabeled data under the current model, the selection strategy corresponding to the next round of autonomous learning is determined; the attribute information is obtained by labeling the unlabeled data according to the current model; wherein, if it is determined according to the validation set indicators and the attribute information that the current autonomous learning does not meet the stage conditions, the first selection strategy is determined as the selection strategy corresponding to the next round of autonomous learning; if it is determined that the current autonomous learning meets the stage conditions, the selection strategy corresponding to the next round of autonomous learning is determined by combining the first selection strategy and the second selection strategy.

8. An electronic device, characterized in that: include: processor; as well as A memory, configured to store executable instructions of the processor; Wherein, the processor is configured to perform the data processing method described in any one of claims 1-6 by executing the executable instructions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the data processing method according to any one of claims 1 to 6 is implemented.

Images